Add project metadata for uv-managed development environments

[bzantium]

Fixes #7090.

Summary

This makes the repository root a valid Python project for tools like Astral uv without changing the existing Bazel-based TensorBoard build/run workflow.

Changes:

• add standard setuptools build metadata at the repo root
• add a PEP 621 project table in pyproject.toml
• source runtime dependencies from tensorboard/pip_package/requirements.txt
• add a tensorflow optional extra that installs tf-nightly
• document the uv workflow in DEVELOPMENT.md
• ignore common local uv artifacts (.venv/, uv.lock)

Motivation

Before this change, the repo root had a pyproject.toml but no project table, so commands like uv sync and uv pip install -e . failed from a fresh checkout.

This change gives contributors a straightforward way to manage the development environment with uv, while still using Bazel to build and run TensorBoard itself.

Verification

I verified:

• uv sync
• uv sync --extra tensorflow
• uv pip install -e .

I did not verify uv run bazel run //tensorboard -- --logdir ... here because Bazel is not installed in this environment.